3 files changed, 184 insertions, 191 deletions

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index d445a64b979e..736461a861b1 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -84,23 +84,71 @@ xfs_find_bdev_for_inode(
 }
 
 /*
- * We're now finished for good with this ioend structure.
- * Update the page state via the associated buffer_heads,
- * release holds on the inode and bio, and finally free
- * up memory.  Do not use the ioend after this.
+ * We're now finished for good with this page.  Update the page state via the
+ * associated buffer_heads, paying attention to the start and end offsets that
+ * we need to process on the page.
+ */
+static void
+xfs_finish_page_writeback(
+	struct inode		*inode,
+	struct bio_vec		*bvec,
+	int			error)
+{
+	unsigned int		end = bvec->bv_offset + bvec->bv_len - 1;
+	struct buffer_head	*head, *bh;
+	unsigned int		off = 0;
+
+	ASSERT(bvec->bv_offset < PAGE_SIZE);
+	ASSERT((bvec->bv_offset & ((1 << inode->i_blkbits) - 1)) == 0);
+	ASSERT(end < PAGE_SIZE);
+	ASSERT((bvec->bv_len & ((1 << inode->i_blkbits) - 1)) == 0);
+
+	bh = head = page_buffers(bvec->bv_page);
+
+	do {
+		if (off < bvec->bv_offset)
+			goto next_bh;
+		if (off > end)
+			break;
+		bh->b_end_io(bh, !error);
+next_bh:
+		off += bh->b_size;
+	} while ((bh = bh->b_this_page) != head);
+}
+
+/*
+ * We're now finished for good with this ioend structure.  Update the page
+ * state, release holds on bios, and finally free up memory.  Do not use the
+ * ioend after this.
  */
 STATIC void
 xfs_destroy_ioend(
-	xfs_ioend_t		*ioend)
+	struct xfs_ioend	*ioend,
+	int			error)
 {
-	struct buffer_head	*bh, *next;
+	struct inode		*inode = ioend->io_inode;
+	struct bio		*last = ioend->io_bio;
+	struct bio		*bio, *next;
 
-	for (bh = ioend->io_buffer_head; bh; bh = next) {
-		next = bh->b_private;
-		bh->b_end_io(bh, !ioend->io_error);
-	}
+	for (bio = &ioend->io_inline_bio; bio; bio = next) {
+		struct bio_vec	*bvec;
+		int		i;
 
-	mempool_free(ioend, xfs_ioend_pool);
+		/*
+		 * For the last bio, bi_private points to the ioend, so we
+		 * need to explicitly end the iteration here.
+		 */
+		if (bio == last)
+			next = NULL;
+		else
+			next = bio->bi_private;
+
+		/* walk each page on bio, ending page IO on them */
+		bio_for_each_segment_all(bvec, bio, i)
+			xfs_finish_page_writeback(inode, bvec, error);
+
+		bio_put(bio);
+	}
 }
 
 /*
@@ -174,7 +222,8 @@ xfs_setfilesize(
 
 STATIC int
 xfs_setfilesize_ioend(
-	struct xfs_ioend	*ioend)
+	struct xfs_ioend	*ioend,
+	int			error)
 {
 	struct xfs_inode	*ip = XFS_I(ioend->io_inode);
 	struct xfs_trans	*tp = ioend->io_append_trans;
@@ -188,53 +237,32 @@ xfs_setfilesize_ioend(
 	__sb_writers_acquired(VFS_I(ip)->i_sb, SB_FREEZE_FS);
 
 	/* we abort the update if there was an IO error */
-	if (ioend->io_error) {
+	if (error) {
 		xfs_trans_cancel(tp);
-		return ioend->io_error;
+		return error;
 	}
 
 	return xfs_setfilesize(ip, tp, ioend->io_offset, ioend->io_size);
 }
 
 /*
- * Schedule IO completion handling on the final put of an ioend.
- *
- * If there is no work to do we might as well call it a day and free the
- * ioend right now.
- */
-STATIC void
-xfs_finish_ioend(
-	struct xfs_ioend	*ioend)
-{
-	if (atomic_dec_and_test(&ioend->io_remaining)) {
-		struct xfs_mount	*mp = XFS_I(ioend->io_inode)->i_mount;
-
-		if (ioend->io_type == XFS_IO_UNWRITTEN)
-			queue_work(mp->m_unwritten_workqueue, &ioend->io_work);
-		else if (ioend->io_append_trans)
-			queue_work(mp->m_data_workqueue, &ioend->io_work);
-		else
-			xfs_destroy_ioend(ioend);
-	}
-}
-
-/*
  * IO write completion.
  */
 STATIC void
 xfs_end_io(
 	struct work_struct *work)
 {
-	xfs_ioend_t	*ioend = container_of(work, xfs_ioend_t, io_work);
-	struct xfs_inode *ip = XFS_I(ioend->io_inode);
-	int		error = 0;
+	struct xfs_ioend	*ioend =
+		container_of(work, struct xfs_ioend, io_work);
+	struct xfs_inode	*ip = XFS_I(ioend->io_inode);
+	int			error = ioend->io_bio->bi_error;
 
 	/*
 	 * Set an error if the mount has shut down and proceed with end I/O
 	 * processing so it can perform whatever cleanups are necessary.
 	 */
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
-		ioend->io_error = -EIO;
+		error = -EIO;
 
 	/*
 	 * For unwritten extents we need to issue transactions to convert a
@@ -244,55 +272,33 @@ xfs_end_io(
 	 * on error.
 	 */
 	if (ioend->io_type == XFS_IO_UNWRITTEN) {
-		if (ioend->io_error)
+		if (error)
 			goto done;
 		error = xfs_iomap_write_unwritten(ip, ioend->io_offset,
 						  ioend->io_size);
 	} else if (ioend->io_append_trans) {
-		error = xfs_setfilesize_ioend(ioend);
+		error = xfs_setfilesize_ioend(ioend, error);
 	} else {
 		ASSERT(!xfs_ioend_is_append(ioend));
 	}
 
 done:
-	if (error)
-		ioend->io_error = error;
-	xfs_destroy_ioend(ioend);
+	xfs_destroy_ioend(ioend, error);
 }
 
-/*
- * Allocate and initialise an IO completion structure.
- * We need to track unwritten extent write completion here initially.
- * We'll need to extend this for updating the ondisk inode size later
- * (vs. incore size).
- */
-STATIC xfs_ioend_t *
-xfs_alloc_ioend(
-	struct inode		*inode,
-	unsigned int		type)
+STATIC void
+xfs_end_bio(
+	struct bio		*bio)
 {
-	xfs_ioend_t		*ioend;
-
-	ioend = mempool_alloc(xfs_ioend_pool, GFP_NOFS);
-
-	/*
-	 * Set the count to 1 initially, which will prevent an I/O
-	 * completion callback from happening before we have started
-	 * all the I/O from calling the completion routine too early.
-	 */
-	atomic_set(&ioend->io_remaining, 1);
-	ioend->io_error = 0;
-	INIT_LIST_HEAD(&ioend->io_list);
-	ioend->io_type = type;
-	ioend->io_inode = inode;
-	ioend->io_buffer_head = NULL;
-	ioend->io_buffer_tail = NULL;
-	ioend->io_offset = 0;
-	ioend->io_size = 0;
-	ioend->io_append_trans = NULL;
+	struct xfs_ioend	*ioend = bio->bi_private;
+	struct xfs_mount	*mp = XFS_I(ioend->io_inode)->i_mount;
 
-	INIT_WORK(&ioend->io_work, xfs_end_io);
-	return ioend;
+	if (ioend->io_type == XFS_IO_UNWRITTEN)
+		queue_work(mp->m_unwritten_workqueue, &ioend->io_work);
+	else if (ioend->io_append_trans)
+		queue_work(mp->m_data_workqueue, &ioend->io_work);
+	else
+		xfs_destroy_ioend(ioend, bio->bi_error);
 }
 
 STATIC int
@@ -364,50 +370,6 @@ xfs_imap_valid(
 		offset < imap->br_startoff + imap->br_blockcount;
 }
 
-/*
- * BIO completion handler for buffered IO.
- */
-STATIC void
-xfs_end_bio(
-	struct bio		*bio)
-{
-	xfs_ioend_t		*ioend = bio->bi_private;
-
-	if (!ioend->io_error)
-		ioend->io_error = bio->bi_error;
-
-	/* Toss bio and pass work off to an xfsdatad thread */
-	bio->bi_private = NULL;
-	bio->bi_end_io = NULL;
-	bio_put(bio);
-
-	xfs_finish_ioend(ioend);
-}
-
-STATIC void
-xfs_submit_ioend_bio(
-	struct writeback_control *wbc,
-	xfs_ioend_t		*ioend,
-	struct bio		*bio)
-{
-	atomic_inc(&ioend->io_remaining);
-	bio->bi_private = ioend;
-	bio->bi_end_io = xfs_end_bio;
-	submit_bio(wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE, bio);
-}
-
-STATIC struct bio *
-xfs_alloc_ioend_bio(
-	struct buffer_head	*bh)
-{
-	struct bio		*bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES);
-
-	ASSERT(bio->bi_private == NULL);
-	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
-	bio->bi_bdev = bh->b_bdev;
-	return bio;
-}
-
 STATIC void
 xfs_start_buffer_writeback(
 	struct buffer_head	*bh)
@@ -452,28 +414,35 @@ static inline int xfs_bio_add_buffer(struct bio *bio, struct buffer_head *bh)
 }
 
 /*
- * Submit all of the bios for an ioend. We are only passed a single ioend at a
- * time; the caller is responsible for chaining prior to submission.
+ * Submit the bio for an ioend. We are passed an ioend with a bio attached to
+ * it, and we submit that bio. The ioend may be used for multiple bio
+ * submissions, so we only want to allocate an append transaction for the ioend
+ * once. In the case of multiple bio submission, each bio will take an IO
+ * reference to the ioend to ensure that the ioend completion is only done once
+ * all bios have been submitted and the ioend is really done.
  *
  * If @fail is non-zero, it means that we have a situation where some part of
  * the submission process has failed after we have marked paged for writeback
- * and unlocked them. In this situation, we need to fail the ioend chain rather
- * than submit it to IO. This typically only happens on a filesystem shutdown.
+ * and unlocked them. In this situation, we need to fail the bio and ioend
+ * rather than submit it to IO. This typically only happens on a filesystem
+ * shutdown.
  */
 STATIC int
 xfs_submit_ioend(
 	struct writeback_control *wbc,
-	xfs_ioend_t		*ioend,
+	struct xfs_ioend	*ioend,
 	int			status)
 {
-	struct buffer_head	*bh;
-	struct bio		*bio;
-	sector_t		lastblock = 0;
-
 	/* Reserve log space if we might write beyond the on-disk inode size. */
 	if (!status &&
-	     ioend->io_type != XFS_IO_UNWRITTEN && xfs_ioend_is_append(ioend))
+	    ioend->io_type != XFS_IO_UNWRITTEN &&
+	    xfs_ioend_is_append(ioend) &&
+	    !ioend->io_append_trans)
 		status = xfs_setfilesize_trans_alloc(ioend);
+
+	ioend->io_bio->bi_private = ioend;
+	ioend->io_bio->bi_end_io = xfs_end_bio;
+
 	/*
 	 * If we are failing the IO now, just mark the ioend with an
 	 * error and finish it. This will run IO completion immediately
@@ -481,33 +450,73 @@ xfs_submit_ioend(
 	 * time.
 	 */
 	if (status) {
-		ioend->io_error = status;
-		xfs_finish_ioend(ioend);
+		ioend->io_bio->bi_error = status;
+		bio_endio(ioend->io_bio);
 		return status;
 	}
 
-	bio = NULL;
-	for (bh = ioend->io_buffer_head; bh; bh = bh->b_private) {
+	submit_bio(wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE,
+		   ioend->io_bio);
+	return 0;
+}
 
-		if (!bio) {
-retry:
-			bio = xfs_alloc_ioend_bio(bh);
-		} else if (bh->b_blocknr != lastblock + 1) {
-			xfs_submit_ioend_bio(wbc, ioend, bio);
-			goto retry;
-		}
+static void
+xfs_init_bio_from_bh(
+	struct bio		*bio,
+	struct buffer_head	*bh)
+{
+	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
+	bio->bi_bdev = bh->b_bdev;
+}
 
-		if (xfs_bio_add_buffer(bio, bh) != bh->b_size) {
-			xfs_submit_ioend_bio(wbc, ioend, bio);
-			goto retry;
-		}
+static struct xfs_ioend *
+xfs_alloc_ioend(
+	struct inode		*inode,
+	unsigned int		type,
+	xfs_off_t		offset,
+	struct buffer_head	*bh)
+{
+	struct xfs_ioend	*ioend;
+	struct bio		*bio;
 
-		lastblock = bh->b_blocknr;
-	}
-	if (bio)
-		xfs_submit_ioend_bio(wbc, ioend, bio);
-	xfs_finish_ioend(ioend);
-	return 0;
+	bio = bio_alloc_bioset(GFP_NOFS, BIO_MAX_PAGES, xfs_ioend_bioset);
+	xfs_init_bio_from_bh(bio, bh);
+
+	ioend = container_of(bio, struct xfs_ioend, io_inline_bio);
+	INIT_LIST_HEAD(&ioend->io_list);
+	ioend->io_type = type;
+	ioend->io_inode = inode;
+	ioend->io_size = 0;
+	ioend->io_offset = offset;
+	INIT_WORK(&ioend->io_work, xfs_end_io);
+	ioend->io_append_trans = NULL;
+	ioend->io_bio = bio;
+	return ioend;
+}
+
+/*
+ * Allocate a new bio, and chain the old bio to the new one.
+ *
+ * Note that we have to do perform the chaining in this unintuitive order
+ * so that the bi_private linkage is set up in the right direction for the
+ * traversal in xfs_destroy_ioend().
+ */
+static void
+xfs_chain_bio(
+	struct xfs_ioend	*ioend,
+	struct writeback_control *wbc,
+	struct buffer_head	*bh)
+{
+	struct bio *new;
+
+	new = bio_alloc(GFP_NOFS, BIO_MAX_PAGES);
+	xfs_init_bio_from_bh(new, bh);
+
+	bio_chain(ioend->io_bio, new);
+	bio_get(ioend->io_bio);		/* for xfs_destroy_ioend */
+	submit_bio(wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE,
+		   ioend->io_bio);
+	ioend->io_bio = new;
 }
 
 /*
@@ -523,27 +532,24 @@ xfs_add_to_ioend(
 	struct buffer_head	*bh,
 	xfs_off_t		offset,
 	struct xfs_writepage_ctx *wpc,
+	struct writeback_control *wbc,
 	struct list_head	*iolist)
 {
 	if (!wpc->ioend || wpc->io_type != wpc->ioend->io_type ||
 	    bh->b_blocknr != wpc->last_block + 1 ||
 	    offset != wpc->ioend->io_offset + wpc->ioend->io_size) {
-		struct xfs_ioend	*new;
-
 		if (wpc->ioend)
 			list_add(&wpc->ioend->io_list, iolist);
-
-		new = xfs_alloc_ioend(inode, wpc->io_type);
-		new->io_offset = offset;
-		new->io_buffer_head = bh;
-		new->io_buffer_tail = bh;
-		wpc->ioend = new;
-	} else {
-		wpc->ioend->io_buffer_tail->b_private = bh;
-		wpc->ioend->io_buffer_tail = bh;
+		wpc->ioend = xfs_alloc_ioend(inode, wpc->io_type, offset, bh);
 	}
 
-	bh->b_private = NULL;
+	/*
+	 * If the buffer doesn't fit into the bio we need to allocate a new
+	 * one.  This shouldn't happen more than once for a given buffer.
+	 */
+	while (xfs_bio_add_buffer(wpc->ioend->io_bio, bh) != bh->b_size)
+		xfs_chain_bio(wpc->ioend, wbc, bh);
+
 	wpc->ioend->io_size += bh->b_size;
 	wpc->last_block = bh->b_blocknr;
 	xfs_start_buffer_writeback(bh);
@@ -803,7 +809,7 @@ xfs_writepage_map(
 			lock_buffer(bh);
 			if (wpc->io_type != XFS_IO_OVERWRITE)
 				xfs_map_at_offset(inode, bh, &wpc->imap, offset);
-			xfs_add_to_ioend(inode, bh, offset, wpc, &submit_list);
+			xfs_add_to_ioend(inode, bh, offset, wpc, wbc, &submit_list);
 			count++;
 		}
 
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index b4421177b68d..814aab790713 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -18,7 +18,7 @@
 #ifndef __XFS_AOPS_H__
 #define __XFS_AOPS_H__
 
-extern mempool_t *xfs_ioend_pool;
+extern struct bio_set *xfs_ioend_bioset;
 
 /*
  * Types of I/O for bmap clustering and I/O completion tracking.
@@ -37,22 +37,19 @@ enum {
 	{ XFS_IO_OVERWRITE,		"overwrite" }
 
 /*
- * xfs_ioend struct manages large extent writes for XFS.
- * It can manage several multi-page bio's at once.
+ * Structure for buffered I/O completions.
  */
-typedef struct xfs_ioend {
+struct xfs_ioend {
 	struct list_head	io_list;	/* next ioend in chain */
 	unsigned int		io_type;	/* delalloc / unwritten */
-	int			io_error;	/* I/O error code */
-	atomic_t		io_remaining;	/* hold count */
 	struct inode		*io_inode;	/* file being written to */
-	struct buffer_head	*io_buffer_head;/* buffer linked list head */
-	struct buffer_head	*io_buffer_tail;/* buffer linked list tail */
 	size_t			io_size;	/* size of the extent */
 	xfs_off_t		io_offset;	/* offset in the file */
 	struct work_struct	io_work;	/* xfsdatad work queue */
 	struct xfs_trans	*io_append_trans;/* xact. for size update */
-} xfs_ioend_t;
+	struct bio		*io_bio;	/* bio being built */
+	struct bio		io_inline_bio;	/* MUST BE LAST! */
+};
 
 extern const struct address_space_operations xfs_address_space_operations;
 
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index a9ea12d46b74..b412bb1c5fd3 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -58,8 +58,7 @@
 #include <linux/parser.h>
 
 static const struct super_operations xfs_super_operations;
-static kmem_zone_t *xfs_ioend_zone;
-mempool_t *xfs_ioend_pool;
+struct bio_set *xfs_ioend_bioset;
 
 static struct kset *xfs_kset;		/* top-level xfs sysfs dir */
 #ifdef DEBUG
@@ -1699,20 +1698,15 @@ MODULE_ALIAS_FS("xfs");
 STATIC int __init
 xfs_init_zones(void)
 {
-
-	xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
-	if (!xfs_ioend_zone)
+	xfs_ioend_bioset = bioset_create(4 * MAX_BUF_PER_PAGE,
+			offsetof(struct xfs_ioend, io_inline_bio));
+	if (!xfs_ioend_bioset)
 		goto out;
 
-	xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
-						  xfs_ioend_zone);
-	if (!xfs_ioend_pool)
-		goto out_destroy_ioend_zone;
-
 	xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
 						"xfs_log_ticket");
 	if (!xfs_log_ticket_zone)
-		goto out_destroy_ioend_pool;
+		goto out_free_ioend_bioset;
 
 	xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
 						"xfs_bmap_free_item");
@@ -1808,10 +1802,8 @@ xfs_init_zones(void)
 	kmem_zone_destroy(xfs_bmap_free_item_zone);
  out_destroy_log_ticket_zone:
 	kmem_zone_destroy(xfs_log_ticket_zone);
- out_destroy_ioend_pool:
-	mempool_destroy(xfs_ioend_pool);
- out_destroy_ioend_zone:
-	kmem_zone_destroy(xfs_ioend_zone);
+ out_free_ioend_bioset:
+	bioset_free(xfs_ioend_bioset);
  out:
 	return -ENOMEM;
 }
@@ -1837,9 +1829,7 @@ xfs_destroy_zones(void)
 	kmem_zone_destroy(xfs_btree_cur_zone);
 	kmem_zone_destroy(xfs_bmap_free_item_zone);
 	kmem_zone_destroy(xfs_log_ticket_zone);
-	mempool_destroy(xfs_ioend_pool);
-	kmem_zone_destroy(xfs_ioend_zone);
-
+	bioset_free(xfs_ioend_bioset);
 }
 
 STATIC int __init